CN1252488C - Magnetic resonance imaging method with sub-sampling - Google Patents

Abstract

A magnetic resonance imaging method is proposed wherein a magnetic resonance image is reconstructed from magnetic resonance signals from respective signal channels. More specifically, individual signal channels relate to respective surface coils which are employed as receiver antennas for the magnetic resonance signals. The magnetic resonance signals are acquired with sub-sampling of the k-space. Resampling on a regular square grid is performed, thus enabling fast Fourier transformation in the reconstruction of the magnetic resonance image. Furthermore, the reconstruction is carried out on the basis of the spatial sensitivity profile of the receiver antennas, i.e. of the surface coils, so as to separate contributions from different spatial positions in the sub-sampled magnetic resonance signals. Preferably, a spiral-shaped acquisition trajectory is followed in the k-space.

Description

The method and system of magnetic resonance imaging

Background technology

The present invention relates to a kind of MR imaging method that forms magnetic resonance image (MRI),

Wherein by receiving antenna by a plurality of signalling channel collecting magnetic resonance signals,

Its independent receiving antenna has sensitivity profile separately.

The invention still further relates to a kind of magnetic resonance system.

Article " Coil Sensitivity Encoding for Fast MRI " (people's such as K.P.Prussmann, deliver on Proceedings ISMRM (1998) 579) discloses a kind of MR imaging method and has implemented the magnetic resonance system of this MR imaging method.

Known MR imaging method has the method for being abbreviated as SENSE.This known MR imaging method is used the receiving antenna of the form of receiving coil.The sub sampling that this MR imaging method is used the magnetic resonance signal of being gathered with reduce to the k-space that is used for required visual field and in the k-space for scanning the k-required time of space with a sampling density on the enough big zone of the required spatial resolution of magnetic resonance image (MRI), it should be noted that carrying out the corresponding line that scans along its in the k-space is provided with more fartherly than being provided with of requisite space resolution needs in the k-space.In other words, in the k-space " line is skipped ".Because this " skipping of k-space center line ", so collecting magnetic resonance signal needs the less time.Rebuild the receiving coil image according to magnetic resonance signal from the sub sampling of independent receiving coil.Because this sub sampling has reduced actual visual field, therefore in this receiving coil image, produced the pseudo-shadow of reverse stack and aliasing.From the receiving coil image, draw magnetic resonance image (MRI) according to sensitivity profile, basic thus or even eliminated the pseudo-shadow of aliasing in this magnetic resonance image (MRI) fully.This no aliasing operation expands magnetic resonance image (MRI) to required visual field.

Have been found that in the radiology practice the required time of collecting magnetic resonance signal needs further to reduce considerably.Have been found that, particularly for the tissue site to rapid movement, for example nervous patient's pulsatile heart is dirty, carries out the MR imaging method of high spatial resolution imaging and for MR angiogram method, need substantially reduce the acquisition time of magnetic resonance imaging

An object of the present invention is to provide a kind of MR imaging method, wherein the acquisition time of magnetic resonance signal is than acquisition time much shorter required when using known SENSE technology.

Realize this purpose by MR imaging method according to the present invention, wherein be illustrated in Noise Correlation between the independent signalling channel by noise correlation matrix, here,

Use the sub sampling collecting magnetic resonance signal,

The rule of sampling out again from the magnetic resonance signal of being gathered on the sampling grid of rule is the magnetic resonance signal of sampling again,

By block diagonal matrix or band diagonal matrix this noise correlation matrix is similar to, the value that is positioned at the matrix element outside the predetermined band around the principal diagonal of approximate noise correlation matrix is zero, and

The magnetic resonance signal of sampling again according to the rule of having carried out sampling again from the magnetic resonance signal of being gathered on the basis of sensitivity profile and approximate noise correlation matrix is rebuild magnetic resonance image (MRI).

When utilizing sensitivity profile, from the magnetic resonance signal of sub sampling the k-space, draw magnetic resonance image (MRI).Sub sampling means that in the k-space sampling is more coarse, promptly the resolution that in the k-space, has than be used for the magnetic resonance image (MRI) visual field enough resolution coarse.In a kind of MR imaging method, the minimum wavelength that the brightness of magnetic resonance image (MRI) changes is relevant with the visual field.Minimum wavelength especially with the visual field be in proportion and proportional with the sampling density in the k-space.Under the situation of sub sampling, the sampling more required than the size of the visual field that enough is used for magnetic resonance image (MRI) of sampling is coarse.Signal value according to they in the k-space wave vector and be carried out coding according to sensitivity profile.The magnetic resonance signal of receiving antenna separately is corresponding to separately signalling channel.The noise that contributes to the signal in each signalling channel comes the linear combination of contribution of the noise of other all signalling channel of autocorrelative signalling channel and (in principle).Receiving antenna is for example to the receiving coil of magnetic resonance signal sensitivity.Preferably, use surface coils as receiving antenna.This surface coils is arranged on the patient's body that will check, and this surface coils picks up near the magnetic resonance signal of position surface coils that produces significantly in the patient's body that will check.Noise between the signalling channel is relevant to be represented with noise correlation matrix.Number to the actual resonance signal of the magnetic resonance image (MRI) that is used for quality of diagnosis, if do not take measures, be that the matrix inversion problem of calculated capacity of having relatively high expectations and the computing time of growing has been caused with the pixel value that magnetic resonance signal is decoded into the independent location of pixels that is used for image array in the basis in the k-space and with the sensitivity profile.The noise relevant issues can be approximate by unit matrix, block diagonal matrix or two diagonal matrix, and all these matrixes all are the specific examples of block diagonal matrix or band diagonal matrix.From the sub sampling magnetic resonance signal, rebuild magnetic resonance image (MRI) according to the SENSE algorithm and comprise the noisiness optimization that makes in magnetic resonance image (MRI).This optimization relates to noise correlation matrix, this noise correlation matrix be included in the diagonal element noise in the magnetic resonance signal of being sampled and the magnetic resonance signal of the corresponding sampling of being gathered by different receiver antennas between the off-diagonal element noise relevant in.Can substitute noise correlation matrix with unit matrix as approximate as can be seen.As a kind of replacement, more approximate being based on the recognition: noise is relevant constant substantially in time.Therefore, showing can be by having sparsity structure, and it is relevant that promptly approximate (piece) cornerwise matrix is described noise fully.This sparsity structure allows the magnetic resonance signal from the sub sampling of separately receiver coil is carried out actual sample or reinstalling in the physical channel as the linear combination from the magnetic resonance signal of the sub sampling of independent receiver coil.To noise correlation matrix the so-called Cholesky of the product of its matrix that becomes reversible left triangular matrix and its close conjugation in distress is decomposed, obtain the weight in this linear combination thus.So, the effective noise correlation matrix that links the physical channel is a unit matrix.Have been found that according to the present invention in practice and to carry out suitably approximate at the better simply matrix that only has from relevant available between the noise contribution in the magnetic resonance signal of independent receiving antenna near the contribution of the element the principal diagonal.Even have been found that this noise is relevant and can replace with unit matrix.Have been found that this simplification has greatly alleviated the matrix inversion problem, therefore only require relatively short computing time and limited calculated capacity.Therefore in the short time cycle, can from magnetic resonance signal, rebuild magnetic resonance image (MRI).Also find, in one minute, can from the sub sampling magnetic resonance signal, rebuild 128 * 128 image arrays (so N=128) in practice.Also have been found that the approximate influence that the quality of diagnosis of magnetic resonance image (MRI) is not had significant adverse according to the noise correlation matrix of sensitivity profile in the reconstruction of magnetic resonance image (MRI).This means that magnetic resonance image (MRI) has suitable contrast resolution, therefore in magnetic resonance image (MRI), visually reproduced low contrast details suitably.By known fast fourier transform being applied to the reconstruction time that has further reduced magnetic resonance image (MRI) in the regular magnetic resonance signal of sampling again significantly.Rule (again) sampling means is sampling to the magnetic resonance signal in the k-space on the square net of rule.Have been found that for N * N image array, this simplification makes the matrix inversion problem from N ⁴The order of magnitude be reduced to N ²Or N ²The order of magnitude of logN.

The invention provides selection to be used for the height freedom of track collection, that following the k-space of magnetic resonance signal.This acquisition trajectories has caused the non-rule sampling in k-space according to the present invention.Especially in the process of the collection of magnetic resonance signal, need in the k-space, not sample by the square net to rule.Therefore, for example, can pass the corresponding part in k-space with different speed.The possibility that the present invention especially provides selection to pass the basic spiral trajectory in k-space.Collecting magnetic resonance signal from the core in k-space at first utilizes the wave vector of relatively little size to this then, thereafter collecting magnetic resonance signal when the size of wave vector increases continuously apace.Spiral helicine track or comprise that this collection of the track of one or more spiral fashion step-lengths is particularly suitable for using in MR angiogram method in the k-space.In this method, applying contrast medium to the patient, forming the patient's that will check magnetic resonance image (MRI) after for example by the intravenous injection contrast medium immediately.Magnetic resonance signal from the center in k-space is chiefly directed to structure quite coarse in the magnetic resonance image (MRI), comprises the artery part of the patient's vascular system that will check.The vein segment of vascular system mainly comprises much trickleer structure.Following in the spiral trajectory, before contrast medium reaches vein from artery part collecting magnetic resonance signal.In addition, because used sub sampling, so the collection of magnetic resonance signal and do not require the long time.The collection of the sub sampling of magnetic resonance signal and make it possible to gather apace the magnetic resonance image (MRI) of the artery part of vascular system along the combination of the scanning of spiral trajectory with higher spatial resolution.

Describe these aspects of the present invention and other aspect in detail according to the embodiment hereinafter that is limited in the dependent claims.

Preferably, from rebuilding separately receiving coil image from independent signalling channel thereby from the magnetic resonance signal the receiving antenna separately.Preferably receiving coil is used as receiving antenna.Owing to, in this receiving coil image, produced the pseudo-shadow of aliasing such as reverse superposition phenomenon from the sub sampling of the magnetic resonance signal of independent signalling channel.Use approximate noise correlation matrix butt joint take-up loop graph according to the present invention looks like to rebuild.Magnetic resonance image (MRI) draws from the receiving coil image according to sensitivity profile.Be called the SENSE method according to receiving coil image and sensitivity profile reconstruction magnetic resonance image (MRI) itself.This SENSE method was open in the article of MRM42 (1999) pp.952-962 at the article of delivering on the Proceedings ISMRM (1998) 579 and Prussmann and Weiger by people such as Prussmann originally.The SENSE method can realize the collection of the sub sampling of magnetic resonance signal, can reduce the required time of collection of magnetic resonance signal thus.

In addition, make up by the magnetic resonance signal to sub sampling, the magnetic resonance signal that forms full sampling from the magnetic resonance signal of sub sampling when utilizing sensitivity profile is possible.Magnetic resonance image (MRI) is rebuild according to the magnetic resonance signal that is obtained by combination.Various magnetic resonance signals in the k-space are combined then to be filled in the line of having skipped in the k-space in gatherer process.This method is known as the abb. SMASH that only gets initial, and itself is from U.S. Pat 5,910,728.

When using receiving coil or surface coils as receiving antenna, the coil sensitivity profiles of receiving coil is corresponding to the sensitivity profile of receiving antenna.

Preferably, receiving coil is preferably basically to the inductance decoupling.Because the inductive coupling degree of receiving coil is lower, so noise level is relevant with noise all lower.Reduced the noise level of magnetic resonance image (MRI) thus.

Preferably, use the iteration inversion algorithms to carry out the reconstruction of magnetic resonance image (MRI).That is, magnetic resonance image (MRI) is rebuild by iteration from the magnetic resonance signal of sub sampling.Begin with a certain initialization vector, iterative algorithm produces the progression of an approximate solution, and this series convergence is to separating accurately.There are a plurality of this technology to handle bigger linear system.So-called conjugate gradient (cg) method is particularly suitable for.On the one hand, be effectively calculating, it can be combined with FFT.On the other hand, CG iteration and do not require and guarantee the convergent particular provisions.If the matrix that is comprised is a positive definite, it can be restrained safely, relates to the matrix of the magnetic resonance signal of sub sampling for the pixel value of the magnetic resonance image (MRI) that will rebuild by gradient coding and coil sensitivity profiles, and this puts and all keeps correct.The CG algorithm is in theory at N at the most ²Produce N after the inferior iteration ²* N ²The exact solution of system.For the N in 128 scope, though its reality is not carried out whole procedure up to realized strict convergence on mathematics.Yet, in practice, after iteration, can obtain to produce magnetic resonance image (MRI) approximate of the reconstruction of quality of diagnosis preferably through less relatively number of times.Each CG iterative step is to comprise matrix to be inverted be multiply by a residual vector and several uncomplicated further calculating.Therefore, the iteration speed key depends on how can carry out the matrix vector multiplication apace.Realize the state of the required iterations of given accuracy and so-called matrix to be inverted and begin the adaptability of vector relevant.Because the dimension and the size of the matrix inversion of the method according to this invention, iteration inversion algorithms ratio such as direct inversion algorithms are faster.For example can obtain particularly advantageous result by Jacobi program, Gauss-Seidel program or conjugate gradient (CG) method.

The invention still further relates to the magnetic resonance imaging system that is suitable for implementing according to MR imaging method of the present invention.In independent claims 8, define according to magnetic resonance imaging system of the present invention.

Magnetic resonance imaging system according to the present invention comprises the control module of the computing machine that has (little) processor, control time gradient fields and RF excitation thus.Preferably computing machine by suitably programming or (little) processor or application specific processor are implemented the function according to magnetic resonance imaging system of the present invention, and this computing machine or (little) processor or application specific processor have specialized designs and be used to carry out integrated electronic or optoelectronic circuit according to one or more MR imaging method of the present invention.

The invention still further relates to a kind of computer program that has the instruction of carrying out MR imaging method.Further purpose of the present invention provides a kind of computer program, can implement thus according to one or more MR imaging method of the present invention.In independent claims 9, limit according to computer program of the present invention.When will computer program loads according to the present invention in the computing machine of magnetic resonance imaging system the time, this magnetic resonance imaging system be carried out according to one or more MR imaging method of the present invention.Therefore, on basis, can realize producing technique effect according to magnetic resonance image (MRI) of the present invention according to the instruction of computer program of the present invention.For example, magnetic resonance imaging system according to the present invention is that a kind of its computing machine is mounted with the magnetic resonance imaging system according to computer program of the present invention.This computer program can be stored in carrier for example among the CD-ROM.Reading computer program (for example passing through cd-rom player) then from this carrier is stored in computer program loads in the storer of computing machine of magnetic resonance imaging system in computing machine and with it.Yet, it should be noted, can also according to network for example World Wide Web will computer program loads according to the present invention in the storer of the computing machine of magnetic resonance imaging system.

With reference to hereinafter and the described embodiment of accompanying drawing, clearly set forth these and other aspect of the present invention by non-limiting instance.

Summary of drawings

Show to this accompanying drawing diagrammatic and use magnetic resonance imaging system of the present invention therein.

Iii. accompanying drawing is described

This accompanying drawing diagrammatic shows and uses magnetic resonance imaging system of the present invention therein.This magnetic resonance imaging system comprises one group of main coil 10, has produced stable, uniform magnetic field thus.For example construct main coil like this: they have surrounded tunnel-shaped inspection space.The patient who checks slides in this tunnel-shaped inspection space.This magnetic resonance imaging system also comprises a plurality of gradient coils 11,12, and the magnetic field that produces the presentation space variation thus especially is the magnetic field of interim gradient profile on independent direction, to be superimposed upon on the uniform magnetic field.Gradient coil 11,12 is connected to controllable power supply unit 21.Applying electric current by power supply unit 21 encourages gradient coil 11,12.Control intensity, direction and the duration of gradient by the control power supply unit.This magnetic resonance imaging system also comprise be respectively applied for produce RF driving pulse and picking up magnetic resonance signals transmit and receive coil 13,16.Transmitting coil 13 preferable configuration are body coil 13, can surround the target that will check (a part) thus.Body coil is set in magnetic resonance imaging system usually like this: body coil 13 can surround the patient 30 that will check when he or she is placed in the magnetic resonance imaging system.Body coil 13 is as the emitting antenna of transmitting RF driving pulse and RF refocusing pulse.Preferably, body coil 13 comprises the uniform strength distribution spatially of the RF pulse (RFS) of being launched.Can use identical coil or antenna replacedly as transmitting coil and receiving coil.In addition, transmitting and receiving the coil general shape is coil, but other the coil that wherein transmits and receives also is feasible as the geometric configuration that transmits and receives antenna of electromagnetic signal, transmits and receives coil 13 and is connected to electronics and transmits and receives on the circuit 15.

It should be noted that in addition, it also is possible using independent receiving coil 16.For example, use surface coils 16 as receiving coil.This surface coils has high sensitivity in less relatively space.Transmitting coil is connected to detuner 24 such as surface coils, carries out demodulation by 24 pairs of magnetic resonance signals that received of detuner (MS).To be applied in the reconstruction unit through the magnetic resonance signal (DMS) of demodulation.Receiving coil is connected to prime amplifier 23.Prime amplifier 23 amplifies the RF resonance signal (MS) that is received by receiving coil 16, and the resonance signal through amplifying is used in the detuner 24.24 pairs of resonance signals through amplifying of detuner carry out demodulation.The actual information that includes the local spin density in the position of wanting imaging that relates at object through the resonance signal of demodulation.In addition, transmit and receive circuit 15 and be connected to modulator 22.Modulator 22 and transmit and receive circuit 15 and start transmitting coils 13 with transmitting RF excitation and refocusing pulse.Reconstruction unit draws one or more picture signals from the magnetic resonance signal (DMS) of institute's demodulation, the image information at the imaging position of the object that this picture signal indicates to check.Preferably reconstruction unit 25 is configured to Digital Image Processing unit 25 in practice, is programmed with the picture signal from the image information of wanting the imaging position that draws indicated object through the magnetic resonance signal of demodulation in this Digital Image Processing unit 25.To be applied on the monitor 26 at the signal in the output of reconstruction unit, so that monitor shows magnetic resonance image (MRI).Interchangeable is that magnetic resonance signal can also be represented three-dimensional Density Distribution.Can in many ways this three-dimensional Density Distribution be presented on the monitor 26, for example show that projection or stereo-picture that the user will select are right.Also interchangeablely be, when products for further is handled can also with from the signal storage of rebuilding unit 25 in buffer cell 27.

According to magnetic resonance imaging system of the present invention also have comprise (little) processor for example with the control module 20 of the form of computing machine.The execution of control module 20 control RF excitations and the application of interim gradient fields.For this reason, for example will computer program loads according to the present invention in control module 20 and reconstruction unit 25.

From the MR imaging method that is known as the SENSE method, can learn, between measured magnetic resonance signal (m) and brightness or correlative value (I), have following relation: I=(E ^HΨ ^-1E) ^-1E ^HΨ ^-1M (1)

Wherein encoder matrix with according to the wave vector of magnetic resonance signal in the k-space space encoding and according to sensitivity profile (s _γ) (r _p) space encoding relevant, subscript γ represents the surface coils of being correlated with, r here _pBe illustrated in the relevant volume element in the object that to check or the position of voxel.E ^HThe multiple adjoint matrix of presentation code matrix.Noise correlation matrix is expressed as Ψ, and it has following matrix element:

${\mathrm{\Ψ}}_{\mathrm{\γ\η}}=\underset{\mathrm{\τ}}{\mathrm{\Σ}}{\mathrm{\ω}}_{\mathrm{\γ\τ}}{{\mathrm{\σ}}_{\mathrm{\τ}}}^2{\mathrm{\ω}}_{\mathrm{\γ\η}}$

Here σ _τBe illustrated in the noise standard deviation among the signalling channel τ, ω _{γ τ}Expression signalling channel τ is to the weighting coefficient of the noise contribution in this signalling channel γ.

Encoder matrix E has following matrix element thus:

$E_{\mathrm{\γ},\mathrm{\κ},\mathrm{\ρ}}=s_{\mathrm{\γ}}\left(r_{\mathrm{\ρ}}\right)e^{{\mathrm{ik}}_{\mathrm{\κ}}{r}_{\mathrm{\ρ}}}---\left(2\right)$

S wherein _γ(r _p) be illustrated in the area r of receiving antenna, particularly surface coils γ _pOn spatial sensitivity profile.According to the present invention, with block diagonal matrix in addition can carry out fully noise correlation matrix Ψ with unit matrix approximate, especially when the decoupling of surface coils inductance especially like this.Then efficient coding is reduced to:

In method more cleverly, the noise of hypothesis between different coils was relevant along with the time is constant when mechanism below was constant.The block diagonal matrix that promptly has simple structure by sparse matrix is described receiver noise then:

${\widetilde{\mathrm{\Ψ}}}_{(\mathrm{\γ},\mathrm{\κ}),({\mathrm{\γ}}^{\′},{\mathrm{\κ}}^{\′})}={\mathrm{\Ψ}}_{\mathrm{\γ},{\mathrm{\γ}}^{\′}}{\mathrm{\δ}}_{\mathrm{\κ},{\mathrm{\κ}}^{\′}}.$

Time, irrelevant matrix Ψ can determine by the statistical study sample plot that the reference noise that is carried out when not having the MR signal is sampled.Use η _γThe noise output of expression the-passage, then the item of Ψ is provided by following formula:

${\mathrm{\Ψ}}_{\mathrm{\γ},{\mathrm{\γ}}^{\′}}=\stackrel{\‾}{{\mathrm{\η}}_{\mathrm{\γ}}{\mathrm{\η}}_{{\mathrm{\γ}}^{\′}}^{*}},$

Here it is average to go up the line express time.In the method, taken into full account the Noise Correlation of the magnetic resonance signal of gathering simultaneously.

Use the noise statistics method of this simplification, just can eliminate the noise variance matrix by simple skill.The linear combination of passing through Src Chan of basic thought produces the receiver channel of one group of reality, does not have mutual Noise Correlation so that actual passage has the unit noise level.Draw suitable weighting coefficient by the inverse matrix of decomposing the matrix L that obtains by Cholesky for this reason:

Ψ＝LL ^H

By following formula, the actual samples data with decorrelation unit noise can obtain from crude sampling thus:

$m_{\mathrm{\γ},\mathrm{\κ}}^{\mathrm{decorr}}=\underset{{\mathrm{\γ}}^{\′}}{\mathrm{\Σ}}{\left(L^{-1}\right)}_{\mathrm{\γ},{\mathrm{\γ}}^{\′}}{m}_{{\mathrm{\γ}}^{\′},\mathrm{\κ}}.$

The clean coil sensitivity of getting in touch with the physical channel provides by following formula:

$s_{\mathrm{\γ}}^{\mathrm{decorr}}\left(r\right)=\underset{{\mathrm{\γ}}^{\′}}{\mathrm{\Σ}}{\left(L^{-1}\right)}_{\mathrm{\γ},{\mathrm{\γ}}^{\′}}{s}_{{\mathrm{\γ}}^{\′}}\left(r\right),$

${\left({\hat{E}}^H\hat{E}\right)}_{\mathrm{\ρ},{\mathrm{\ρ}}^{\′}}=\underset{\mathrm{\γ}}{\mathrm{\Σ}}{\hat{s}}_{\mathrm{\γ}}^{*}\left(r_{\mathrm{\ρ}}\right){\hat{s}}_{\mathrm{\γ}}\left(r_{{\mathrm{\ρ}}^{\′}}\right)(\∫e^{-\mathrm{ik}(r_{\mathrm{\ρ}}-r_{{\mathrm{\ρ}}^{\′}})}\left(\underset{\mathrm{\κ}}{\mathrm{\Σ}}\mathrm{\δ}(k_{\mathrm{\κ}}-k)\right)\mathrm{dk})$

Obtain modified encoder matrix

$E_{(\mathrm{\γ},\mathrm{\κ}),\mathrm{\ρ}}^{\mathrm{decorr}}=e^{{\mathrm{ik}}_{\mathrm{\κ}}{r}_{\mathrm{\ρ}}}{s}_{\mathrm{\γ}}^{\mathrm{decorr}}\left(r_{\mathrm{\ρ}}\right).$

Utilize sampled value and the sensitivity revised by this way, can handle the passage of combination recently as handling a physical channel.Therefore the noise variance matrix of physical channel equals identical, can save it when explaining equation [6] again when carrying out image reconstruction from modified data.It should be noted that by transforming to actual passage, separating of equation [6] do not change (seeing appendix B).Specifically, the optimization that has kept SNR.

On the one hand, carry out decorrelation, and on the other hand, it is relevant to ignore noise, chooses for these two kinds to have caused identical formula of reduction.Remove the subscript that is used for decorrelation, still nearly like encoder matrix be expressed as the E that it all reads in two kinds of situations (formula more above (3)):

Represent the encoder matrix that is similar to now by benchmark E.Have been found that by the iteration numerical value ground of inverting apace and solve this matrix inversion problem.Data m is the signal value (amplitude and phase place) of magnetic resonance signal.Sign indicating number in (3) comprises inverse Fourier transform, therefore

Wherein Ω is the general expression of residue code, to be different from inverse Fourier transform.This can be written as vector and matrix element more accurately:

${\left(\hat{E}y\right)}_{(\mathrm{\γ},\mathrm{\κ})}=\∫e^{{-\mathrm{ik}}_{\mathrm{\κ}}r}\left(\underset{\mathrm{\ρ}}{\mathrm{\Σ}}{y}_{\mathrm{\ρ}}{\hat{s}}_{\mathrm{\γ}}\left(r_{\mathrm{\ρ}}\right)\mathrm{\δ}(r_{\mathrm{\ρ}}-r)\mathrm{dr}\right)$

${\left({\hat{E}}^{H}m\right)}_{\mathrm{\ρ}}=\underset{\mathrm{\γ}}{\mathrm{\Σ}}{\hat{s}}_{\mathrm{\γ}}^{*}\left(r_{\mathrm{\ρ}}\right)(\∫e^{{-\mathrm{iknr}}_{\mathrm{\ρ}}}\left(\underset{\mathrm{\κ}}{\mathrm{\Σ}}{m}_{(\mathrm{\γ},\mathrm{\κ})}\mathrm{\δ}(k_{\mathrm{\κ}}-k)\right)\mathrm{dk})$

And

For carrying out inverse Fourier transform apace, in the k-space, on the square net of rule, measured data are carried out so-called " gridding " conversion:

M=G  m, G for example is a Gaussian convolution nuclear here:

${\tilde{m}}_{\mathrm{\κ}}=\mathrm{\ΣG}(\mathrm{\κ}-{\mathrm{\κ}}^{\′})m\left({\mathrm{\κ}}^{\′}\right)---\left(5\right)$

Therefore:

Wherein can carry out inverse Fourier transform apace by known fast Fourier transform (FFT) algorithm.The FFT that uses Kaiser-Bessel window and twice (two-fold) over-sampling to utilize gridding to prepare calculates the integration in these expression formulas effectively.The assessing the cost of each equation of estimating these equatioies only is N ²The order of magnitude of logN, and conventional matrix vector multiplication is N ⁴The order of magnitude.The function f that is stored in the formula [5] requires about N ²Memory-size rather than directly the storage

Required N ⁴For

Double counting, advantageously the method with gridding is divided into dual mode:

A kind of is at first to calculate and store integration and carry out common matrix vector multiplication, creates line by line And do not store them.The advantage of this program be it and common method than other restrain Gauss-Seidel method compatibility faster.On the other hand, the complexity of a matrix vector multiplication still remains on N substantially ⁴

Another kind is to carry out by two continuous griddings and FFT step Calculating.The advantage of this method is that the cost of matrix vector multiplication only is N ²The order of magnitude of logN, and this program is carried out on parallel processing hardware easily.In addition, invert by iteration and from the residue code among sign indicating number Ω, rebuild magnetic resonance image (MRI).In general, this iteration is inverted and can be written as: I ⁽ⁱ⁺¹⁾=I ⁽ⁱ⁾+ α Δ (m ⁽ⁱ⁾, m ^(i-1)), signal data m here ⁽ⁱ⁾Be view data I on the basis of residue code Ω from being rebuild ⁽ⁱ⁾Middle calculating.Each new estimation I that on the basis of different function Δs, forms view data ⁽ⁱ⁺¹⁾, this new estimation is each closer to be fit to by the measured signal data of magnetic resonance signal.Iteration is with view data I ⁽⁰⁾Estimate and measured magnetic resonance signal m ⁽⁰⁾Beginning.

Claims (10)

1. MR imaging method that forms magnetic resonance image (MRI), wherein

By receiving antenna by a plurality of signalling channel collecting magnetic resonance signals,

This independent receiving antenna all has sensitivity profile separately, and

The noise that is illustrated between the independent signalling channel with noise correlation matrix is relevant, wherein

Utilize the sub sampling collecting magnetic resonance signal,

The rule of sampling out again from the magnetic resonance signal of being gathered on the sampling grid of rule is the magnetic resonance signal of sampling again,

With block diagonal matrix or band diagonal matrix noise correlation matrix is similar to, the value that is positioned at the matrix element outside the predetermined band around the principal diagonal of the noise correlation matrix that be similar to is zero, and

The magnetic resonance signal of sampling again according to the rule of having carried out sampling from the magnetic resonance signal of being gathered on the basis of sensitivity profile and approximate noise correlation matrix is rebuild magnetic resonance image (MRI).

2. the described MR imaging method of claim 1, wherein Jin Si noise correlation matrix is a diagonal matrix.

3. the described MR imaging method of claim 2, wherein, approximate noise correlation matrix is a unit matrix.

4. the described MR imaging method of claim 1, wherein

Rebuild corresponding receiving coil image according to the magnetic resonance signal of sampling again from the rule of independent signalling channel, and

Magnetic resonance image (MRI) draws from receiving coil image and sensitivity profile.

5. the described MR imaging method of claim 1, the decoupling substantially each other of the sensitivity profile of wherein independent receiving antenna.

6. each described MR imaging method of claim 1-5, wherein

Magnetic resonance image (MRI) is the magnetic resonance signal of sampling again according to rule, rebuilds by the iteration inversion algorithms.

7. the described MR imaging method of claim 6, wherein magnetic resonance image (MRI) is according to the receiving coil image, rebuild by the iteration inversion algorithms.

8. each described MR imaging method of claim 1-5, wherein magnetic resonance signal is to gather on the basis of the track by the k-space, and this track is corresponding to the sampling of the magnetic resonance signal of being gathered outside the regular grid of the sampled point in the k-space.

9. the described MR imaging method of claim 8, the track that wherein passes the k-space comprises spiral helicine substantially part.

10. a magnetic resonance system is used to form magnetic resonance image (MRI), and this system is configured to:

By receiving antenna and through a plurality of signalling channel collecting magnetic resonance signals,

This independent receiving antenna all has sensitivity profile separately, and

The noise that is illustrated between the independent signalling channel with noise correlation matrix is relevant, wherein

Utilize the sub sampling collecting magnetic resonance signal,

The rule of sampling out again from the magnetic resonance signal of being gathered on the sampling grid of rule is the magnetic resonance signal of sampling again,

With block diagonal matrix or band diagonal matrix noise correlation matrix is similar to, the value that is positioned at the matrix element outside the predetermined band around the principal diagonal of the noise correlation matrix that be similar to is zero, and

The magnetic resonance signal of sampling again according to the rule of having carried out sampling from the magnetic resonance signal of being gathered on the basis of sensitivity profile and approximate noise correlation matrix is rebuild magnetic resonance image (MRI).

Images